I'm seriously considering to mod my K340 to drive directly the electret drivers.

I would really appreciate some comments on the following calculations though.

Here is the network inside the cans:


DKK32 is the 600ohms dynamic driver
R1 is 680ohms iirc
u58 is the step-up transformer, measured to have a 8:1 ratio by Fitz ( http://www.head-fi.org/forums/f4/k-3...ml#post3249621 )
R2 is omitted in the most recent versions, let's forget it.

The overall impedance being 360ohms, it means that the primary of u58 is of roughly 220ohms. 900ohms (the 680 of R1+ 220) in parallel with the 600ohms of DKK32 would give us the nominal 360ohms.

The weird thing is that it would mean that the signal at the electret drivers is barely twice the input voltage (R1 and u58 acting as a 1/4 voltage divider) ?? The impedance ratio of the transformer being the square of the turn ratio, we'd also have a secondary load of 220*64= 14K. Seems low for an electret.

Or do I have to understand from Fitz' post that the electret receive a 8x signal ? That'd mean that u58 has a 32:1 ratio, giving us an impedance of 225k for the electret. It seems somehow more plausible.

.

Wow, this is quite a serious endeavor. You have my respect, sir!

The (former) user aerius mentioned the idea of biamping the K340 as well. dunno if he ever got around to doing it.

R1 and U58 dont act as a voltage divider exactly the way you describe. 220ohms is the DC resistance of the primary of the transformer, the AC impedance could be in the several K-ohm ranges. As fitz said, the voltage stepup is slightly less than 1:8 with the resistor in place.

There should be a few ways to bi-amp the 340's, but I think that R1 in series with the transformer like that as well as the high output impedance of the secondary will probably roll off the high frequency output of the electret a bit stock. I think that biamping will change this quite a bit, perhaps to the point where the electret has no roll-off.

I wonder if anyone has experimented with better transformers mounted outside of the headphones. That would be a fun project. Even changing R1 to a different value could show big changes if your amp can drive it.

As another guess, a long tail pair with a push-pull 10K:150 OPT with the electret cap coupled to the plates and the dynamic driver off of the transformer could be really sweet if you dont listen too loud. Perhaps a gain stage in front if you do listen loud

Quote:



*smacks head* it serves me well for posting at 2am. You're right of course.

I went through the documentation available... it's pretty difficult to get an idea of the capacitive load of the electret drivers.

My idea of bi-amplifying is using the electrostatic amp currently developped by Runeight for the electret and yet-to-be decided tube otl amp for the dynamic driver.

I attempted to model the K340. Here is what it looks like for now in LTspice:



V1 is the input signal.
R1 is the 600ohm dynamic driver (which certainly doesn't behave so IRL)
R2 is the serie resistor
R3 is the DC impedance of the winding (good luck guessing how it changes with frequency, it's pretty much irrelevant though and I could have let it out)
F1 is a current controlled current source. It simulates the load placed on the system by the electret through the transformer. To do so, it multiplies by 8 the current going through C1 (for transformer current gain).
E1 is a voltage controlled voltage source. It provides the 8x voltage gain of the transformer.
R5 is the output impedance of the transformer. I took the square of the gan ratio (64), multiplied by R2. It's probably worse than that.
C1 is the electret driver. I took a very heavy load, no idea what it could be in reality.
V3 is necessary for ltspice's CCCS but discard it.

E1, R5 and C1 don't interact directly with the rest of the system (only through the CCCS). But getting a xformer model to work here was too messy.

Conclusions:
- this thing is 360ohms at DC, 600ohms until 1K and slowly goes down to 450ohms at 20k. This could explain why this headphones are so difficult to drive, they really would be 600ohms and not 360ohms. More voltage drive to provide than expected...
- The voltage swing for the electret is seriously getting limited as the frequency goes up. From a 200Vp-p swing @10khz to 110Vp-p @20khz.
- As expected, the high frequency response might suffer from the output impedance of the xformer.

If anyone has an idea on how to improve the model (or correct it, or explain why it's bad), I'll gladly take it.